CN111242276A - One-dimensional convolution neural network construction method for load current signal identification - Google Patents

One-dimensional convolution neural network construction method for load current signal identification Download PDF

Info

Publication number
CN111242276A
CN111242276A CN201911377565.0A CN201911377565A CN111242276A CN 111242276 A CN111242276 A CN 111242276A CN 201911377565 A CN201911377565 A CN 201911377565A CN 111242276 A CN111242276 A CN 111242276A
Authority
CN
China
Prior art keywords
neural network
load
dimensional
construction method
convolutional neural
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201911377565.0A
Other languages
Chinese (zh)
Other versions
CN111242276B (en
Inventor
陈伟
赵国伟
李慧
姚志芳
姚泽宁
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
North China Electric Power University
Datong Power Supply Co of State Grid Shanxi Electric Power Co Ltd
Original Assignee
North China Electric Power University
Datong Power Supply Co of State Grid Shanxi Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by North China Electric Power University, Datong Power Supply Co of State Grid Shanxi Electric Power Co Ltd filed Critical North China Electric Power University
Priority to CN201911377565.0A priority Critical patent/CN111242276B/en
Publication of CN111242276A publication Critical patent/CN111242276A/en
Application granted granted Critical
Publication of CN111242276B publication Critical patent/CN111242276B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Business, Economics & Management (AREA)
  • Economics (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Neurology (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a one-dimensional convolution neural network construction method for load current signal identification, which comprises the following steps: step S1: collecting user load data as a training set and a test set; step S2: building a one-dimensional convolution neural network model; step S3: training a built one-dimensional convolutional neural network identification model through a training set; step S4: and inputting the test set into the trained one-dimensional convolutional neural network for recognition to obtain a load recognition result. The method is improved on the basis of the traditional convolutional neural network model, and the load identification efficiency is improved by adopting the one-dimensional convolutional neural network through reducing the algorithm complexity; sliding on the signal using the convolution kernel as an observation window ensures the time-varying and frequency-band correlation of the signal. The time sequence characteristics are automatically extracted through the convolution kernel, and the accuracy rate of load identification is improved.

Description

One-dimensional convolution neural network construction method for load current signal identification
Technical Field
The invention belongs to the technical field of power load identification, and particularly relates to a one-dimensional convolution neural network construction method for load current signal identification.
Background
The non-invasive load monitoring mode is that a monitoring tool which is convenient to install and maintain and high in efficiency is used for feeding back total power utilization data to a data processing center and transmitting the total power utilization data to a power grid in real time, so that the power grid can obtain the power configuration and consumption of residential users, and a more reasonable power transmission and distribution scheme is formulated. With the diverse needs of life and the diverse development of the kinds of electric appliances, the research of non-invasive load monitoring is also being deepened and expanded.
In the traditional intrusive load monitoring method, a hardware sensor is arranged at each load, and collected electricity utilization information is transmitted to a processing center through carrier transmission. The method has the advantages of accurate measurement and the defects that when the number of the electric equipment in the monitored system is large, the traditional load detection mode is high in cost and inconvenient in installation of the sensor device and system maintenance. The non-invasive load detection does not need to install a sensor at each monitored load, but extracts the electricity consumption data of all electric appliances on the service main line, and the influence on the monitored equipment is minimized, so that the phase application electricity information is extracted. However, the existing monitoring facilities do not have the functions of storage and self-learning for monitoring the power load, and bring difficulty to non-invasive power load monitoring.
Disclosure of Invention
In order to solve the technical problems, the invention solves the problems of difficult monitoring of non-invasive power loads and poor self-learning function, and realizes the intelligent technical effect.
Specifically, the invention provides a one-dimensional convolution neural network construction method for load current signal identification, which is used for load current signal identification, and is characterized in that the neural network construction method comprises the following steps:
step S1: collecting user load data as a training set and a test set;
step S2: building a one-dimensional convolution neural network model;
step S3: training a built one-dimensional convolutional neural network identification model through a training set;
step S4: and inputting the test set into the trained one-dimensional convolutional neural network for recognition to obtain a load recognition result.
Further, in step S1, the user load data is a plurality of kinds of single load current data, and the current signal of the current at the current moment and the previous moment and the newly input single load current can be represented as:
Figure BDA0002341399800000021
where I (t) is the current at the present moment, I' (t) is the current at the previous moment, Im(t) is single load current which is newly put into operation, n (t) is noise, and m represents the number of electrical appliances in the home of a user.
Further, in step S1, the user load data is calculated by a method of 2: a ratio of 1 into the training set and the test set.
Further, in step S2, the one-dimensional convolution operation maps the sequence data in the one-dimensional convolution neural network into a convolution layer, and the generated feature map is:
Figure BDA0002341399800000022
f(z)＝max(z，0)
in the formula: x represents input sequence data, represents a one-dimensional convolution operation,
Figure BDA0002341399800000023
representing by a convolution kernel
Figure BDA0002341399800000024
Generated j characteristic graphs, j is belonged to [1, n ∈ ]c]，ncAnd (b) representing the number of convolution kernels, b is offset, and f (z) is an activation function, wherein the activation function adopts ReLu function acceleration model convergence.
Further, in step S2, the generated feature map is maximum value sampled, and the maximum value sampling model may be represented as:
Figure BDA0002341399800000031
wherein ,
Figure BDA0002341399800000032
is the largest sample value in the feature map and 2k is the feature map length.
Further, in step S2, when the pooling operation is performed for the last time in the one-dimensional convolutional neural network, the global maximum pooling is used to capture the most useful global timing information, and the length of the feature map is reduced to 1; the global max-pooling model may be represented as:
Figure BDA0002341399800000033
in the formula ：
Figure BDA0002341399800000034
representing the last maximum sample value obtained.
Further, in step S2, the output of the classification result of the full-connected layer in the one-dimensional convolutional neural network may be represented as:
Figure BDA0002341399800000035
Figure BDA0002341399800000036
wherein ,WfcIs the weight of the full connection layerMatrix, afcIs a full connection layer time sequence characteristic sequence, an activation function fσ(z) is sigmoid function, z is output sequence of full connection layer, and the value range of z is (- ∞, + ∞); finally outputting classification results
Figure BDA0002341399800000037
Representing the probability of belonging to different categories.
The invention has the beneficial effects that:
compared with the traditional machine learning, the load identification efficiency of the one-dimensional convolutional neural network is improved by reducing the algorithm complexity; sliding on the signal using the convolution kernel as an observation window ensures the time-varying and frequency-band correlation of the signal. The time sequence characteristics are automatically extracted through the convolution kernel, and the accuracy rate of load identification is improved.
Drawings
Fig. 1 is a schematic flowchart of a method for constructing a one-dimensional convolutional neural network for load current signal identification according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a non-intrusive load identification model of a one-dimensional convolutional neural network in a one-dimensional convolutional neural network construction method for load current signal identification according to an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is further specifically described by the following embodiments and the accompanying drawings 1-2.
As shown in fig. 1, an embodiment of the present application provides a method for constructing a one-dimensional convolutional neural network for load current signal identification, which includes the following steps:
step S1: collecting user load data as a training set and a test set;
step S2: building a one-dimensional convolution neural network model;
step S3: training a built one-dimensional convolutional neural network identification model through a training set;
step S4: and inputting the test set into the trained one-dimensional convolutional neural network for recognition to obtain a load recognition result.
In step S1, a sensor is installed on the user electricity meter to collect user voltage, current and power information. And then decomposing the acquired mixed data by using a load decomposition algorithm to obtain single load data and storing the single load data, wherein each load data is used as a training set and a test set.
Specifically, according to the practical operation habits of the user on the loads, the possibility that the two loads are absolutely and synchronously turned on or off at the same time is very low, and time differences of different sizes inevitably exist, that is, the load switching is performed sequentially. Thus, the current at the present moment and the current signal at the previous moment and the newly applied single load current can be expressed as:
Figure BDA0002341399800000041
where I (t) is the current at the present moment, I' (t) is the current at the previous moment, Im(t) is single load current which is newly put into operation, n (t) is noise, and m represents the number of electrical appliances in the home of a user.
By solving the underdetermined equation, two paths of signals are separated at each output time interval, one path is a load current signal which is newly put into operation at present, and the other path is a mixed current signal at the previous moment, and newly-put load data I are obtainedm(t) storing as individual load data. And constructing a load database by using the single load data, carrying out event detection, load decomposition, load clustering and load identification by using prior data such as living habits on the load data, storing each load data, and establishing the load database. The individual load data of all the signals of a subscriber form a dedicated load database for the subscriber. Meanwhile, the load database updates the database in real time according to the newly added electric appliances of the user. Although sampling ensures real-time updating of the load signature database, in real life, once the load database is built, little new load is added in a short time. Thus, the load database will be updated periodically each day, and this process will not be performed at other times to save time.
And (3) the load data in the load database is divided into 2: 1 into a training set and a test set; the training set contains load type information, and the test set does not contain load type information.
In step 2, a one-dimensional convolutional neural network model for load identification is constructed.
The input of the one-dimensional convolution neural network is a one-dimensional vector, so that the convolution kernel and the characteristic diagram inside the network are also one-dimensional. The sequence data is mapped into a convolution layer through one-dimensional convolution operation, and a generated characteristic diagram is as follows:
Figure BDA0002341399800000051
f(z)＝max(z，0)
in the formula: x represents input sequence data, represents a one-dimensional convolution operation,
Figure BDA0002341399800000052
representing by a convolution kernel
Figure BDA0002341399800000053
Generated j characteristic graphs, j is belonged to [1, n ∈ ]c]，ncAnd (3) representing the number of convolution kernels (a plurality of convolution kernels can generate a plurality of characteristic graphs which are connected in parallel to form a convolution layer), b is offset, f (z) is an activation function, and the activation function adopts a ReLu function to accelerate model convergence and enhance sparse representation of the model.
The pooling layer only needs to perform mean sampling or maximum sampling on the one-dimensional feature map of the previous layer. The pooling operation may halve the sequence length and the maximum pooling model may be expressed as:
Figure BDA0002341399800000054
wherein ,
Figure BDA0002341399800000055
is the largest sample value in the feature map and 2k is the feature map length.
When the pooling operation is adopted for the last time, the global maximum pooling is adopted, the most useful global time sequence information is captured, and the length of the feature graph is reduced to 1. The global max-pooling model may be represented as:
Figure BDA0002341399800000061
wherein ,
Figure BDA0002341399800000062
representing the last maximum sample value obtained.
When the fully connected layer splices the previous layer, the feature graph does not need to be reset into a one-dimensional vector any more, because in the one-dimensional network structure, all the feature graphs are one-dimensional structures and are directly connected in front and back, the fully connected layer can further combine the global time sequence features as follows:
afc＝f(ap,lastWfc+b)
wherein ,WfcWeight matrix of full connection layer, afcIs a full connection layer timing signature sequence. The activation function f (z) is likewise ReLu. For the binary classification problem, the number of output units is 1, and the output of the classification result through the full connection layer can be expressed as:
Figure BDA0002341399800000063
Figure BDA0002341399800000064
activation function fσ(z) is a sigmoid function,
Figure BDA0002341399800000065
for the classification result, z is the output sequence of the fully-connected layer, and the range of z is (— infinity, + ∞). Finally outputting classification results
Figure BDA0002341399800000066
Representing the probability of belonging to different categories.
In step S3, the training set data obtained in step S1 is input into the built neural network for training, and parameters of the neural network model are continuously adjusted iteratively through a loss function until the accuracy is greater than a set threshold. And obtaining a relatively perfect one-dimensional convolutional neural network load identification model according to the output higher identification accuracy.
In step S4, the test set data obtained in step S1 is input into a trained one-dimensional convolutional neural network for recognition, and finally a load recognition result is obtained.
Although the present invention has been described in terms of the preferred embodiment, it is not intended that the invention be limited to the embodiment. Any equivalent changes or modifications made without departing from the spirit and scope of the present invention also belong to the protection scope of the present invention. The scope of the invention should therefore be determined with reference to the appended claims.

Claims (7)

1. A one-dimensional convolution neural network construction method for load current signal identification is characterized by comprising the following steps:
step S1: collecting user load data as a training set and a test set;
step S2: building a one-dimensional convolution neural network model;
step S3: training a built one-dimensional convolutional neural network identification model through a training set;
step S4: and inputting the test set into the trained one-dimensional convolutional neural network for recognition to obtain a load recognition result.
2. The neural network construction method according to claim 1, wherein in step S1, the user load data are a plurality of kinds of single load current data, and the current signal of the current and previous time and the newly input single load current are represented as:
Figure FDA0002341399790000011
where I (t) is the current at the present moment, I' (t) is the current at the previous moment, Im(t) is single load current which is newly put into operation, n (t) is noise, and m represents the number of electrical appliances in the home of a user.
3. The neural network construction method according to claim 1, wherein in step S1, the user load data is divided into the training set and the test set according to a ratio of 2: 1.
4. The neural network construction method of claim 1, wherein in step S2, the sequence data in the one-dimensional convolutional neural network is mapped into convolutional layers by a one-dimensional convolutional operation, and the generated feature map is:
Figure FDA0002341399790000012
f(z)＝max(z，0)
in the formula: x represents input sequence data, represents a one-dimensional convolution operation,
Figure FDA0002341399790000013
representing by a convolution kernel
Figure FDA0002341399790000014
Generated j characteristic graphs, j is belonged to [1, n ∈ ]c]，ncAnd (b) representing the number of convolution kernels, b is offset, and f (z) is an activation function, wherein the activation function adopts ReLu function acceleration model convergence.
5. The neural network construction method according to claim 4, wherein in step S2, the generated feature map is maximum value sampled, and the maximum value sampling model is expressed as:
Figure FDA0002341399790000021
wherein ,
Figure FDA0002341399790000022
is the largest sample value in the feature map and 2k is the feature map length.
6. The neural network construction method according to claim 5, wherein in step S2, when the pooling operation is performed for the last time in the one-dimensional convolutional neural network, the global maximum pooling is used to capture the most useful global timing information, and the length of the feature map is reduced to 1; the global max-pooling model may be represented as:
Figure FDA0002341399790000023
in the formula ：
Figure FDA0002341399790000024
representing the last maximum sample value obtained.
7. The neural network construction method according to claim 6, wherein in step S2, the output of the classification result of the full-link layer in the one-dimensional convolutional neural network is represented as:
Figure FDA0002341399790000025
Figure FDA0002341399790000026
wherein ,WfcWeight matrix of full connection layer, afcIs a full connection layer time sequence characteristic sequence, an activation function fσ(z) is sigmoid function, z is output sequence of full connection layer, and the value range of z is (- ∞, + ∞); finally outputting classification results
Figure FDA0002341399790000027
Represents a genusIn different categories of probabilities.
CN201911377565.0A 2019-12-27 2019-12-27 One-dimensional convolutional neural network construction method for load current signal identification Active CN111242276B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911377565.0A CN111242276B (en) 2019-12-27 2019-12-27 One-dimensional convolutional neural network construction method for load current signal identification

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911377565.0A CN111242276B (en) 2019-12-27 2019-12-27 One-dimensional convolutional neural network construction method for load current signal identification

Publications (2)

Publication Number Publication Date
CN111242276A true CN111242276A (en) 2020-06-05
CN111242276B CN111242276B (en) 2023-08-25

Family

ID=70865600

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911377565.0A Active CN111242276B (en) 2019-12-27 2019-12-27 One-dimensional convolutional neural network construction method for load current signal identification

Country Status (1)

Country Link
CN (1) CN111242276B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112435142A (en) * 2020-12-16 2021-03-02 北京航空航天大学 Power load identification method and load power utilization facility knowledge base construction method thereof
CN113393121A (en) * 2021-06-15 2021-09-14 贵州电网有限责任公司 Non-invasive load identification method based on load power fingerprint characteristics
CN113469228A (en) * 2021-06-18 2021-10-01 国网山东省电力公司淄博供电公司 Power load abnormal value identification method based on data flow space-time characteristics
CN113607068A (en) * 2021-07-19 2021-11-05 华中科技大学 Method for establishing and extracting recognition model of photoacoustic measurement signal characteristics

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018171109A1 (en) * 2017-03-23 2018-09-27 北京大学深圳研究生院 Video action detection method based on convolutional neural network
US20180357542A1 (en) * 2018-06-08 2018-12-13 University Of Electronic Science And Technology Of China 1D-CNN-Based Distributed Optical Fiber Sensing Signal Feature Learning and Classification Method
CN109325395A (en) * 2018-04-28 2019-02-12 二十世纪空间技术应用股份有限公司 The recognition methods of image, convolutional neural networks model training method and device
CN110135227A (en) * 2018-02-09 2019-08-16 电子科技大学 A kind of laser point cloud outdoor scene automatic division method based on machine learning
CN110398369A (en) * 2019-08-15 2019-11-01 贵州大学 A kind of Fault Diagnosis of Roller Bearings merged based on 1-DCNN and LSTM
CN110458189A (en) * 2019-06-29 2019-11-15 天津大学 Compressed sensing and depth convolutional neural networks Power Quality Disturbance Classification Method
CN110597240A (en) * 2019-10-24 2019-12-20 福州大学 Hydroelectric generating set fault diagnosis method based on deep learning

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018171109A1 (en) * 2017-03-23 2018-09-27 北京大学深圳研究生院 Video action detection method based on convolutional neural network
CN110135227A (en) * 2018-02-09 2019-08-16 电子科技大学 A kind of laser point cloud outdoor scene automatic division method based on machine learning
CN109325395A (en) * 2018-04-28 2019-02-12 二十世纪空间技术应用股份有限公司 The recognition methods of image, convolutional neural networks model training method and device
US20180357542A1 (en) * 2018-06-08 2018-12-13 University Of Electronic Science And Technology Of China 1D-CNN-Based Distributed Optical Fiber Sensing Signal Feature Learning and Classification Method
CN110458189A (en) * 2019-06-29 2019-11-15 天津大学 Compressed sensing and depth convolutional neural networks Power Quality Disturbance Classification Method
CN110398369A (en) * 2019-08-15 2019-11-01 贵州大学 A kind of Fault Diagnosis of Roller Bearings merged based on 1-DCNN and LSTM
CN110597240A (en) * 2019-10-24 2019-12-20 福州大学 Hydroelectric generating set fault diagnosis method based on deep learning

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
朱锡祥等: "基于一维卷积神经网络的车载语音识别研究", 《微电子学与计算机》 *
顾鑫等: "基于深度CNN和ELM的滚动轴承故障诊断研究", 《兵器装备工程学报》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112435142A (en) * 2020-12-16 2021-03-02 北京航空航天大学 Power load identification method and load power utilization facility knowledge base construction method thereof
CN113393121A (en) * 2021-06-15 2021-09-14 贵州电网有限责任公司 Non-invasive load identification method based on load power fingerprint characteristics
CN113469228A (en) * 2021-06-18 2021-10-01 国网山东省电力公司淄博供电公司 Power load abnormal value identification method based on data flow space-time characteristics
CN113607068A (en) * 2021-07-19 2021-11-05 华中科技大学 Method for establishing and extracting recognition model of photoacoustic measurement signal characteristics
CN113607068B (en) * 2021-07-19 2022-08-05 华中科技大学 Method for establishing and extracting recognition model of photoacoustic measurement signal characteristics

Also Published As

Publication number Publication date
CN111242276B (en) 2023-08-25

Similar Documents

Publication Publication Date Title
CN111242276A (en) One-dimensional convolution neural network construction method for load current signal identification
Hu et al. Federated region-learning: An edge computing based framework for urban environment sensing
CN108899892A (en) One kind being based on CNN non-intrusion type power load decomposition method
CN110288113A (en) A kind of non-intrusion type load intelligent identifying system
CN103020459A (en) Method and system for sensing multiple-dimension electric utilization activities
CN110136024A (en) Obtain the method and device with electrical feature, consumption habit and electricity demand forecasting value of user
CN106680621A (en) Method for identifying residential load electricity utilization based on current signal separation
CN115100466A (en) Non-invasive load monitoring method, device and medium
CN117172601A (en) Non-invasive load monitoring method based on residual total convolution neural network
CN115130788B (en) Big data analysis-based power utilization prediction system
CN116714469A (en) Charging pile health monitoring method, device, terminal and storage medium
CN116298576A (en) Time-segmentation-considered extensible non-invasive load monitoring method
Zhao et al. Anomaly detection and classification of household electricity data: a time window and multilayer hierarchical network approach
CN113673846B (en) Demand situation awareness method under electric power Internet of things architecture
CN115902557A (en) Switch cabinet fault diagnosis processing method and device and nonvolatile storage medium
Liu et al. Wind power prediction based on LSTM-CNN optimization
CN115545240A (en) Method, system, equipment and medium for diagnosing abnormal line loss of low-voltage distribution network transformer area
CN115062686A (en) Multi-KPI (Key performance indicator) time sequence abnormity detection method and system based on multi-angle features
CN114861993A (en) Regional photovoltaic power generation prediction method based on federal learning and deep neural network
CN114462306A (en) Non-invasive electric load decomposition method based on variable weight time domain convolution network
CN114186631A (en) Load identification method based on non-invasive intelligent terminal
Wang et al. Real-time fault diagnosis method for low-voltage power line carrier communication network based on network topology
CN114662832A (en) Population mobility analysis method and device and electronic equipment
Liu et al. Nonintrusive Load Disaggregation Combining with External Attention Mechanism and Seq2Piont
CN113971371B (en) Terminal power grid user equipment identification and safety control method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant